Deep learning models approach for maritime radar target data classification

Abstract

In maritime ra đa systems, reflection signals play a crucial role in target identification. The application of machine learning techniques, such as recurrent neural networks (RNNs) and convolutional neural networks (CNNs), has gained the attention of researchers in the field of ra đa data analysis. Both theoretical and experimental results demonstrate that these techniques can enhance ra đa target classification performance by utilizing a diverse amount of target data. However, the limited availability of real ra đa data has constrained the development of ra đa data analysis techniques. In this paper, we focus on analyzing and evaluating the performance of classification models, including SCNet, TARAN, TACNN, and RFRAN. We conduct experiments and fine-tune several parameters to improve classification performance. Experiments were carried out on Doppler ra đa and maritime ra đa datasets. The results show that SCNet and RFRAN can be optimized to effectively assist in maritime radar target recognition.